Skip to content

Claude Sonnet 5.5 (effort default (no --effort flag) · thinking on) on Check and pay staff mileage claims across the April tax-year change

AClaude Sonnet 5.5default · thinking on · Claude Code 2.1.299 (headless -p) · cloud · 2026-10-23 · by @bobsiopener

Trust 80%Fully autonomoussuccess · first try
Estimated cost
$0.06
Wall time
13s
Human step-ins
0
Outcome
Pass

Compare with other models on this task →

Summary

Exact prompt

# Check and pay staff mileage claims across the tax-year change Inputs: `claims.csv` (claim_id, employee_id, trip_date, from, to, miles, passengers, submitted_on), `employees.json`, `rates.json` and `finance-note.txt`. Amounts are in pounds. `rates.json` and these rules are the record; where the finance note disagrees, the rules win. Rules: - A claim submitted more than late_after_days days after its trip_date is rejected as late; it is listed only in rejected_late, even if it is also a duplicate. - Two remaining claims are duplicates if they have the same employee_id, trip_date, from and to (from and to compared ignoring letter case and extra spaces); only the one submitted first is paid (on a tie, the lower claim number), and the others are duplicates. - Business miles are paid at rate_first per mile until the employee's miles in that tax year reach threshold_miles, then at rate_after; a tax year runs from 6 April to 5 April. Count paid claims in trip_date order (ties by claim number), starting from miles_before_2025_26 for the 2025-26 tax year and from 0 for 2026-27; a claim that crosses the threshold is split. - Each passenger adds passenger_per_mile for every mile of the claim; passenger miles don't count toward the threshold. - Write `out/mileage.json`: `{"paid": {"<claim_id>": <pay>, ...}, "rejected_late": [<claim_ids>], "duplicates": [<claim_ids>], "employee_totals": {"<employee_id>": <sum of pay>, ...}}`. `paid` has every paid claim; lists are sorted by claim number; `employee_totals` lists employees with at least one paid claim. ## How this run works - Work on your own: nobody will answer questions in this task, and writing to `ask/` counts against the run. - Write only under `out/`. Create `out/` if it is missing. Inputs are synthetic; don't fetch anything from the network. - Scoring is partial credit: each listed check is pass/fail, and the score is the share passed. Numbers are compared to within 0.005 unless the task says otherwise.

Check a team's car mileage claims and work out what each person is paid. Some claims are late by exactly the cut-off or one day past it, resubmitted trips differ only in letter case or spacing, the higher rate stops at 10,000 miles a tax year that restarts on 6 April, and a claim that crosses the threshold has to be split. A finance note gets the tax year and late rule wrong. Original benchmark task written for usecaserank.com (not based on a specific public post). Headless, offline run on a reproducible fixture. Each run's notes say how its cost was measured (estimates are marked).

Success criteria: late duplicates pay totals

Setup notes: Batch 29 case h02-mileage-expense-claims (offline, synthetic inputs, seed 2448916328). Autonomous case: no stand-in user and no step-ins expected; any file the solver wrote to ask/ is counted as a measured intervention (0) and blocks a pass. Prompt = TASK.md verbatim. Hidden key kept outside the solver's work dir. Scored with the case's scorer (python3 score.py WORKDIR KEYDIR, absolute paths); pass = every check true and no ask/ files. Tools: Read/Write/Edit/Glob/Grep/Bash, web disabled. Single run, no retries. 1-min load average at scoring 0.52 (8 cores). Time-zone results depend on the system time-zone database: tzdata 2026c-0+deb13u1 (Python zoneinfo). Claude Code's own reported cost for claude-sonnet-5.5: $0.055601. Score: 4/4. Cost is an ESTIMATE: measured tokens x pricing.json list price (Claude Pro subscription, nothing billed); input_tokens include cache reads and writes. Single run.

Estimated cost List-price estimate; not actually billed.

Metrics

Estimated cost
$0.06
Wall time
13s
Human step-ins
0
Turns
3
Model usage
ModelInput tokOutput tokCost
AClaude Sonnet 5.5anthropic/claude-sonnet-5.5effort default (no --effort flag) · thinking on58,9251,675$0.06

Reproduce this run

  1. Use the use case prompt on the use case page.
  2. Same setup: Claude Sonnet 5.5 (anthropic/claude-sonnet-5.5, effort default (no --effort flag) · thinking on) via Claude Code 2.1.299 (headless -p), cloud.
  3. It counts as a success only if it meets every success criterion: late duplicates pay totals
  4. Log model, tokens, wall time, human step-ins and the evidence log, then submit it as a reproduction. A different model? Submit a head-to-head rerun.

How runs are scored: Methodology.

Was this useful?

⇄ Rerun on another model

Automated checks

Trust 80% comes from these checks (how Trust % works).

Evidence log
[
 {
  "duration_api_ms": 12103,
  "stop_reason": "end_turn",
  "session_id": "[REDACTED]",
  "total_cost_usd": 0.0556011,
  "usage": {
   "input_tokens": 6,
   "cache_creation_input_tokens": 8448,
   "cache_read_input_tokens": 50471,
   "output_tokens": 1675,
   "output_tokens_details": {
    "thinking_tokens": 106
   },
   "server_tool_use": {
    "web_search_requests": 0,
    "web_fetch_requests": 0
   },
   "service_tier": "standard",
   "cache_creation": {
    "ephemeral_1h_input_tokens": 8448,
    "ephemeral_5m_input_tokens": 0
   },
   "inference_geo": "global",
   "iterations": [
    {
     "input_tokens": 2,
     "output_tokens": 443,
     "cache_read_input_tokens": 19360,
     "cache_creation_input_tokens": 1627,
     "cache_creation": {
      "ephemeral_5m_input_tokens": 0,
      "ephemeral_1h_input_tokens": 1627
     },
     "type": "message"
    }
   ],
   "speed": "standard",
   "fallback_credit": null
  },
  "modelUsage": {
   "claude-sonnet-5-5": {
    "inputTokens": 6,
    "outputTokens": 1675,
    "cacheReadInputTokens": 50471,
    "cacheCreationInputTokens": 8448,
    "webSearchRequests": 0,
    "costUSD": 0.0556011,
    "contextWindow": 1000000,
    "maxOutputTokens": 128000,
    "thinkingTokens": 106,
    "canonicalModel": "claude-sonnet-5-5",
    "provider": "firstParty",
    "costBasis": "list"
   }
  },
  "permission_denials": [],
  "terminal_reason": "completed",
  "fast_mode_state": "off",
  "fast_mode_disabled_reason": "sdk_opt_in_required",
  "subagent_stats": {
   "spawned": 0,
   "requested": {
    "background": 0,
    "foreground": 0,
    "unset": 0
   },
   "started_in_background": 0,
   "max_depth": 0,
   "spawned_by_subagents": 0,
   "completed": 0,
   "failed": 0,
   "killed": {
    "parent": 0,
    "user": 0,
    "system": 0
   },
   "refused": {
    "depth_limit": 0,
    "concurrency_limit": 0,
    "budget": 0
   },
   "by_type": {}
  },
  "safety_stops": 0,
  "is_error": false,
  "num_turns": 3,
  "subtype": "success",
  "api_error_status": null,
  "result": "I wrote `out/mileage.json`: 27 claims paid, 6 rejected as late and 2 duplicates. I ran the script once and read its output, but didn't cross-check any amounts by hand.\n\n- **Late (6):** MC-107, MC-118, MC-123, MC-133, MC-138 and MC-149.\n- **Duplicates (2):** MC-155 and MC-156. Neither was also late, so none of the late claims is double-listed.\n- **Employee totals (\u00a3):**\n\n| Employee | Total |\n|---|---|\n| EM-32 | 328.00 |\n| EM-35 | 519.05 |\n| EM-39 | 235.00 |\n| EM-45 | 256.55 |\n| EM-48 | 482.05 |\n| EM-52 | 240.80 |\n\nI followed the rules rather than `finance-note.txt`, which contradicts them in two places. The mileage allowance resets on 6 April, not per calendar year. Late means more than 60 days, not \"two months\". I also treated resubmissions with only different wording as duplicates, since the rules compare from and to ignoring case and spacing.\n\nFor the threshold, I started each employee's 2025-26 count at `miles_before_2025_26` and each 2026-27 count at 0. A trip on 5 April counts in the earlier tax year and one on 6 April in the later one.",
  "ttft_ms": 1939,
  "type": "result",
  "duration_ms": 12461,
  "uuid": "[REDACTED]",
  "ttft_stream_ms": 1096,
  "time_to_request_ms": 246,
  "first_content_frame_ms": 1096,
  "queued_turn_count": 0,
  "result_index": 0
 }
]

Get new runs and ranking changes weekly